Mining gene network by combined association rules and genetic algorithm

Abstract

As gene expression datasets become larger and larger, mining the interaction between genes from large scale gene expression data will be rendered difficult extremely. In this paper, we proposed to combine association rules with genetic algorithm (GA) to mine gene networks from global gene expression data. Firstly, the association rules algorithm is improved from the limitation of the frequent itemsets, the compression of the transaction database and the stored forms of the records in the transaction database for decreasing the computational complexity of this algorithm: Then, the mined rules are further optimized by the genetic algorithm. The efficient combination between two algorithms is realized through the reasonable design of coding/decoding, fitness function and genetic manipulation in GA. An optimizing selection operator is introduced in order to enhance the searching efficiency of GA. Finally, the computational tests have showed that the method combined by association rule and genetic algorithm can mine some important interactions between genes from global gene expression datasets.